AI in Algorithmic Thinking: Creative Problem Generator

Abstract

This study presents a comprehensive architectural roadmap for a purpose-built AI system designed to generate creative, multiple-choice algorithmic and computational thinking problems. Recognizing that effective computational puzzles require narrative logic and visual reasoning rather than rote calculation, the proposed system operates on a “Creativity via Constraints” philosophy. It utilizes a collaborative, multi-agent AI framework—comprising an Architect, an Artist, and a Skeptic—to conceptualize, visualize, and rigorously validate puzzles in a continuous feedback loop. The core architecture features a curated knowledge base that maps fundamental algorithmic topics, alongside a Chain of Thought (CoT) Generator Module that decouples abstract logic from engaging narrative wrappers to formulate algorithmically plausible distractors. Furthermore, a hybrid Visualizer Simulator Module ensures graphical accuracy on an interactive problem solving platform. To guarantee and maintain high educational standards, the framework relies heavily on a Human-in-the-Loop (HITL) validation pipeline driven by educator review dashboards, iterative annotator feedback, and rigorous regression testing. Ultimately, this system provides a scalable, reliable tool that augments educators with high-quality, original question sets while immersing students in engaging problem-solving scenarios.